This paper establishes theoretical guarantees for in-context learning of linear systems using linear transformers, introducing task diversity as a key condition for out-of-domain generalization.

Frank Cole, Yulong Lu, Wuzhe Xu et al.

This study systematically compares 13 radar point-cloud extraction algorithms for ICP odometry in autonomous driving, with K-strongest outperforming others by ~14-25%.

Elliot Preston-Krebs, Daniil Lisus, Timothy D. Barfoot

Proposes a bi-objective ant colony algorithm for robot path planning, maximizing expected reward and survival probability in hazardous environments.

Cory M. Simon, Jeffrey Richley, Lucas Overbey et al.

D-ATACOM method addresses long-term safety and uncertainty with learnable constraints, outperforming existing methods.

Jonas Günster, Puze Liu, Jan Peters et al.

This study evaluates the impact of logging policy bias on two-tower ULTR models using Baidu dataset, finding bias conditions met but limited effect on performance.

Morris de Haan, Philipp Hager

This paper introduces an uncertainty-aware GNN framework for high-level spatial concept generation, improving indoor SLAM accuracy by over 20%.

Jose Andres Millan-Romera, Muhammad Shaheer, Miguel Fernandez-Cortizas et al.

Moshi employs hierarchical Transformer and residual vector quantization for real-time full-duplex speech dialogue with 160ms latency, advancing end-to-end voice systems.

Alexandre Défossez, Laurent Mazaré, Manu Orsini et al.

OmniGen is a unified diffusion model supporting multi-task image generation, including text-to-image, editing, and subject-driven tasks, with simplified architecture and knowledge transfer.

Shitao Xiao, Yueze Wang, Junjie Zhou et al.

Evaluated six Transformer models' music understanding using triplets generated from Audioset ontology.

Yannis Vasilakis, Rachel Bittner, Johan Pauwels

Promptriever is the first instruction-trained dense retriever, achieving state-of-the-art results with nearly 500k instruction-augmented pairs.

Orion Weller, Benjamin Van Durme, Dawn Lawrie et al.

Leveraging large code models for conversational no-code programming of cobots; evaluated on industrial assembly tasks with 85% accuracy in instruction sequences.

Chalamalasetti Kranti, Sherzod Hakimov, David Schlangen

KALE system enhances artwork image captioning with heterogeneous graph, significantly boosting CIDEr scores.

Yanbei Jiang, Krista A. Ehinger, Jey Han Lau

Playground v3 leverages deep fusion with decoder-only LLMs (Llama3-8B) to achieve state-of-the-art text-image alignment, significantly improving prompt adherence.

Bingchen Liu, Ehsan Akhgari, Alexander Visheratin et al.

TRC Bench evaluates 19 LLMs and organizes RAG trustworthiness into six dimensions.

Yujia Zhou, Wenbo Zhang, Jingying Shao et al.

BNEM employs bootstrapped energy matching with diffusion models, reducing variance and improving robustness in high-dimensional sampling.

RuiKang OuYang, Bo Qiang, José Miguel Hernández-Lobato

Proposed multimodal framework integrating motion, audio, and appearance features, achieving state-of-the-art performance on the ARGO1M dataset.

Cagri Gungor, Adriana Kovashka

Proposes the LoCU framework, systematically analyzing vision-language tasks and datasets in comics, highlighting data scarcity and task ambiguity.

Emanuele Vivoli, Mohamed Ali Souibgui, Andrey Barsky et al.

Proposes ManiDext, a diffusion-based framework with continuous correspondence embeddings and residual guidance for realistic hand-object manipulation synthesis.

Jiajun Zhang, Yuxiang Zhang, Liang An et al.

DEAHES-O combines dynamic weighting, EASGD, and AdaHessian, nearly matching oracle fault-aware training on MNIST while beating fixed-weight baselines.

Yuesheng Xu, Arielle Carr

This study shows sparse neural networks (SNNs) can match or outperform dense models on hard samples, especially with limited data and high difficulty, using specific sparsification strategies.

Qiao Xiao, Boqian Wu, Lu Yin et al.